The Debugger

DeepFit – AI sports talent assessment from a single camera India has huge untapped sporting talent, but fitness trials are run by hand with a stopwatch and a measuring tape. They are slow, inconsistent and happen only in a few academies and camps. DeepFit turns any phone or laptop camera into a fair, verified talent-testing station – wherever an athlete is. • 16 camera fitness tests (sit-ups, vertical jump, dumbbell curls, height & weight and more) run on one shared exercise engine. It counts reps from joint angles and rejects half reps. • Trust: the athlete's face is checked before and during every test. The server recomputes every score from the uploaded pose data instead of trusting the phone, and each result is signed (HMAC-SHA256) into a "Talent Passport" with a QR code. A coach scans it and sees VERIFIED, or NOT VALID if anything was edited. • DeepFit Arena: a 3-minute camera game. Dodging a kabaddi raider, jumping a rolling log and reading a cricket bowler secretly measure reflexes, power and anticipation. At the end the player gets a talent profile and the top 3 sports to try. • Heart rate from the face: no watch or sensor – the camera reads tiny skin-colour changes to estimate pulse and heart-rate recovery. • Hidden potential: a Coachability Challenge measures how fast an athlete learns, and growth estimates flag late bloomers who would fail a one-day trial. • Inclusive by design: blind mode speaks every step, and an "audio mirror" tone guides posture by sound. Deaf mode uses captions and colour flashes. The AI coach works in English, Hindi and Telugu. • Meta AI throughout: Muse Spark powers the AI coach, the Arena game master and the photo reading. SAM 3.1, Muse Voice Transcribe and Muse Image add camera calibration, voice commands and a personalised sports poster. Stack: Flask, MongoDB Atlas, MediaPipe, OpenCV, Flutter (offline app), Meta Model API, deployed on Render.